Skip to main content

Duke Investing $1.5 Million in Research Computing

Some of new annual investment available to faculty as vouchers

Duke Professor Craig Henriquez presents a
Duke Professor Craig Henriquez presents a "lightning talk" about cluster computing during the Duke Research Computing Symposium earlier this year.

When Duke professor Craig Henriquez crunches the data for his computer model of the human heart, a single heartbeat takes 12 hours of computing time.

“Given that the heart beats about 100,000 times a day, 36 million times a year, that’s a long time to get one heartbeat,” Henriquez said.

Read More

Storing and analyzing all that data is a challenge far beyond the capabilities of an ordinary computer, so Henriquez relies on a network of more than 450 high-end machines managed by the Office of Information Technology called the Duke Compute Cluster.

While the average desktop or laptop comes with two to four cores (essentially the ‘brains’ of the computer) in its central processing unit and four gigabytes of memory, the Duke Compute Cluster has 4,600 cores on machines that boast 8 to 512 gigabytes of memory each.

Henriquez is one of about 800 Duke researchers who currently tap into the cluster, which allows them to do months of computation in days, or even hours.

Beginning this year, Duke plans to triple its annual investment in research computing, bringing it to roughly $1.5 million annually.

“Computing power is no longer a fancy add-on that’s only for a small cadre of researchers,” Provost Sally Kornbluth told an audience of about 160 people at the 2015 Duke Research Computing Symposium in January. “It’s become an essential tool in all areas of research, from analyzing large social science datasets and complex problems in engineering, to handling data from biological imaging experiments, to solving 3D protein structures. And even in the humanities like the digital reconstruction of ancient cities, for example. It’s everywhere.”

The researchers’ approach illustrates just how far this type of computing has come.

In the early 1960s, many Duke researchers shared a single IBM 7072 to solve their equations. An enormous machine with tape drives and punch cards that whirred and clicked, it was the only digital computer on campus at the time.

“My iPhone is about 100,000 times faster,” Henriquez said.

Some of the new investment in computing will be available to faculty as vouchers, redeemable for use on the Duke Compute Cluster. Other services range from consulting and education to data storage, visualization and management. This includes a new protected data network that offers enhanced security for researchers who work with sensitive data such as health or personnel records.

“The goal is for 80 percent of researchers’ computing and data storage needs to be covered by the university, at no extra cost to researchers, in much the same way that we subsidize other essential support services like the library or the lights,” Kornbluth said.